NLP is imperfect. By definition. It’s impossible to be always right. That’s why one of the fingerprints we are most proud about is our Evaluation Methodology.

It’s opensource, for anyone who wants to use it, here.

In the meantime, we continue to be surprised by the foolish things our system decides to say from time to time:

  • 2-year old babies with type 2 diabetes.
  • Patients with 8 different types of neoplasia.
  • Women with prostate cancer.
  • Patients with >15 knee surgeries or bone fractures.
  • Schrodinger smokers: patients that are both smokers and non-smokers (and even ex-smokers) at the same time.
  • Patients with supposed plenty of medical records… after their exitus.
  • Patients >120 years old.
  • Confusing CI, cardiopatía isquémica and coitus interruptus.
  • Escherechia Coli read as Dr. Coli.
  • Patients that changed gender… >50 times.
  • Kids with dementia.
  • Pregnant men.
  • A patient that supposedly passed away, a couple of days later had diarrhea, fever and then sadly passed away, again.

And more to come...

Thinking about joining our team?

Say hello to us in the following email:


We inform you that the personal data you provide to Medsavana, S.L. with address at Larra 12, ground floor left, 28004, Madrid, will be processed to consider your application in the selection processes in which you decide to participate, as well as to keep you informed of vacancies that may fit your profile. By sending us your CV you consent to the processing of your personal data for the purpose described above. Also, if the selection process in which you participate so requires, your personal data will be transferred to any of the companies of the Savana Group solely to manage the process and your application, if such transfer constitutes an international data transfer, Medsavana will undertake all the appropriate safeguards required from a privacy perspective.
Your personal data will be kept by Medsavana until the end of the corresponding selection process and in any case up to a maximum period of 6 months, as long as you are not hired.
After these periods the data will be deleted.
At any time, you may exercise your rights of access, rectification, erasure, limitation, opposition, as well as the right to data portability and not to be subject to automated individual decisions, by sending a request to the email , providing a copy of your ID card or official document that identifies you. You may also file a complaint with the Spanish Data Protection Agency if you deem it appropriate.

Our Story

Episode I

ICD codes were created in 1898.

Seriously. The coding systems we use around the world in healthcare are from a time when radios or electric fridges didn’t exist. One of our founders got so obsessed about this idea that he left the hospital where he worked as a neurologist, to prove about how incredibly inaccurate and obsolete traditional data approaches are for studying disease.

Then he sat down with his two cofounders in a kitchen. They took a magic-marker and their own medical records and started underlining concepts they wanted to extract.

At the time, they had never heard the Big Data, Natural Language Processing or Machine Learning.

At the time, they had never heard the Big Data, Natural Language Processing or Machine Learning.

But they saw the AI revolution coming. They had an “aha moment” that machine learning, applied to medical data would be a game changer, and that Spain was a great place to start because it was among the countries with a highest penetration of electronic medical records. It still is today.

Ultimately, they realized that, instead of focusing on healthcare “apps” or hospital management tools, they could be the first to apply NLP to real world evidence generation.

Ultimately, they realized that, instead of focusing on healthcare “apps” or hospital management tools, they could be the first to apply NLP to real world evidence generation.

And they had a logo, which was the ugliest logo ever.

And they had a first name, which was also a big mistake

First, they tried to embed their work into a piece of Software called Savana Consulta. And they tried to convince some hospitals that they should use it. But the user experience was terrible, so they failed badly. And during the drive back home they decided to kill that product and focus on NLP and data curation. It was 2015 and that was a great decision.

Episode II

And then, “they” turned into “us”.

We began to redefine ourselves and started partnering with researchers and healthcare providers who shared our vision of a data-driven medicine. At the time, many people thought our mission was unrealistic – nothing less than a global project to drive the use of AI in healthcare.

We understood why people saw it as a kind of pipe dream: an audacious goal only within the reach of large organizations with millions of dollars in funding and a huge marketing budgets. 

We were a small startup that nobody knew and with just enough resources. But we also knew that if we could get enough people involved, we would really impact clinical research. Maybe it was a crazy dream, but nobody could deny that it was a beautiful dream.

Then, more and more companies and researchers began to join us and all that healthy ambition about putting our two cents in leaving a better world, began to bloom. It was a new turning point.

Episode III

But there was still one piece missing. Our Natural Language Processing was a technology, but Science demands more than this; on top of our tech we had to build a methodology. And so did we.
Our team is now combining data scientists and clinicians with experience in clinical and translational research (lead by oncologists).

Today we work with hospitals in countries we have never visited. It may sound normal after COVID, but we have been doing this since 2014, both with customers and with employees, who have never been in our AI Lab in Madrid.

That is great, because we can have amazing talent while they keep enjoying the US Great Lakes or a lasagna on an Italian Island.

We believe in a data-driven Medicine.

We believe in a data-driven Medicine.

We are not the only agents of this historical change – we don’t want to be. Instead, we are happy to belong to a growing community of professionals who work to make AI in healthcare a genuine reality. Savana is definitely not for those who want to generate evidence by the old ways:  filling out registries by hand and using logistic regression to make inferences.  

We send a weekly email for you to start making friends with AI. 

It's not for those who want to generate evidence by traditional means, filling registries by hand and using logistic regression. This is for those who want to leverage machine learning in order to generate evidence in a more automated way and with higher granularity of variables.

Subscribe here:

I don't want to subscribe, I want to read the blog

By the way, if you register you will receive our masterclass about AI in healthcare:

  • 54 min of content. 
  • An astonishing simple explanation about what AI and machine learning are. (Minute 0)
  • Specific and tangible examples of its application in healthcare. (Minute 10)
  • The exact steps to follow in order to start a project in AI. (Minute 21)
  • How to critically read the projects of others. (Minute 34)
  • A tip to distinguish bullshit from truth when buying AI. (Minute 50)
  • A gift: a compelling newsletter which you subscribe to in order to be continually updated, providing you with a quick 5 minute overview just once a week. 
  • Another gift: the best more extensive course to follow providing you with an insightful deeper dive into AI applications in healthcare.

Enrol in a project